Quiz 4 of 5: Workloads, probes and Services
10
On which platform must the team itself schedule and test etcd backups?
Why: GKE and EKS run and back up the control plane for you. On RKE2 or kubeadm clusters the control plane is yours, so etcd snapshots, off-node copies and restore rehearsals are your responsibility.
11
A one-time database migration is deployed as a Deployment and is restarted forever after it succeeds. Which object should it be?
Why: A Deployment's Pods use restartPolicy Always, so a clean exit looks like a crash. A Job understands that exit code zero means the work is done.
12
Which probe failure removes a Pod from Service endpoints without restarting its container?
Why: A failing readiness probe makes the endpoint controller drop the Pod's IP from the EndpointSlice, so traffic stops while the container keeps running and can recover. A failing liveness probe restarts it.